Enhancing Learning in Robotics Teleoperation: Understanding Competence through Eye Fixation Data

Abstract

Aims: A previous study, addressing the analysis of the work activity of a pilot teleoperating a ground robot out-of-sight, led to the hypothesis that the time taken to gather information on the control screens could be correlated with the pilots’ level of skills. This study has made it possible to analyze this hypothesis. Study Design: The study was longitudinal. The activity of the participants was analyzed individually. Place and Duration of Study: Robotic Accident Response Group, Groupe d'intervention robotique sur accidents (INTRA Group, France), between 2021 and 2025. Methodology: Each participant had to perform the same standard robotic teleoperation activity, equipped with an eye-tracking system in order to characterize the pilot's intake of information during the activity. Performance levels were also measured. The standard robotic teleoperation activity consisted of carrying a cylindrical container using the gripper of the robot’s arm whilst moving through a maze, placing the container on the ground halfway, picking up a ring with the gripper, dropping the ring into the container, and then picking the container back up to exit the maze. Results: The data was used to construct an index of competence for each pilot. The consistency of this index of competence with observations made elsewhere has made it possible to validate a scale of competence (\(\tau\)k=.89, p<.005). The construction of a mathematical function involving the pilots’ maximum fixation time and the reference fixation time xref, along with the pilot's fixation time dispersion coefficient cd was correlated with the index of competence (r=.89, p<.001). Conclusion: The study made it possible to measure a pilot's competence in teleoperation based on performance indicators and mastery of the equipment. A model has been proposed to explain how the fixation time during information acquisition on screens correlates with the pilot's competence. The acquisition times for information by a pilot are reduced with the increase in competence and the dispersion of these acquisition times tighten. Thus, training must promote quick scrutiny of all screens: this helps the pilot to get more information about the environment and to build a better mental 3D representation of the robot in the environment.